I’ve been lucky as I haven’t made any really huge mistakes, but if you don’t bring at least part of production down sometimes, you’re not a real sysadmin doing anything consequential.
In my career, I’ve:
1) Rebooted the wrong server (at least twice, prod instead of dev/test).
2) Locked myself out of a firewall, had to drive on site (luckily only 20 minutes and it was after hours) and connect to the console to fix it.
3) Nuked a DHCP scope that was still valid, meaning those computers had nowhere to renew their leases when it came time, keeping them from accessing the network.
Those are just screw-ups I can remember. I am sure there have been others. But as a sysadmin, ultimately even if you test something you’re making the actual change in prod. And for us, prod is never exactly like the test environment as that’s just not possible. And no one is perfect. These things will happen. Having quality backups and a plan to revert to a working known good config is key.